- 18 May, 2018 22 commits
-
-
git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linuxDavid S. Miller authored
Saeed Mahameed says: ==================== mlx5-updates-2018-05-17 mlx5 core dirver updates for both net-next and rdma-next branches. From Christophe JAILLET, first three patche to use kvfree where needed. From: Or Gerlitz <ogerlitz@mellanox.com> Next six patches from Roi and Co adds support for merged sriov e-switch which comes to serve cases where both PFs, VFs set on them and both uplinks are to be used in single v-switch SW model. When merged e-switch is supported, the per-port e-switch is logically merged into one e-switch that spans both physical ports and all the VFs. This model allows to offload TC eswitch rules between VFs belonging to different PFs (and hence have different eswitch affinity), it also sets the some of the foundations needed for uplink LAG support. ==================== Signed-off-by:David S. Miller <davem@davemloft.net>
-
David S. Miller authored
Eric Dumazet says: ==================== tcp: implement SACK compression When TCP receives an out-of-order packet, it immediately sends a SACK packet, generating network load but also forcing the receiver to send 1-MSS pathological packets, increasing its RTX queue length/depth, and thus processing time. Wifi networks suffer from this aggressive behavior, but generally speaking, all these SACK packets add fuel to the fire when networks are under congestion. This patch series adds SACK compression, but the infrastructure could be leveraged to also compress ACK in the future. v2: Addressed Neal feedback. Added two sysctls to allow fine tuning, or even disabling the feature. v3: take rtt = min(srtt, rcv_rtt) as Yuchung suggested, because rcv_rtt can be over estimated for RPC (or sender limited) ==================== Signed-off-by: -
Eric Dumazet authored
This per netns sysctl allows for TCP SACK compression fine-tuning. This limits number of SACK that can be compressed. Using 0 disables SACK compression. Signed-off-by:
Eric Dumazet <edumazet@google.com> Acked-by:
Neal Cardwell <ncardwell@google.com> Signed-off-by:
-
Eric Dumazet authored
This per netns sysctl allows for TCP SACK compression fine-tuning. Its default value is 1,000,000, or 1 ms to meet TSO autosizing period. Signed-off-by:
-
Eric Dumazet authored
This counter tracks number of ACK packets that the host has not sent, thanks to ACK compression. Sample output : $ nstat -n;sleep 1;nstat|egrep "IpInReceives|IpOutRequests|TcpInSegs|TcpOutSegs|TcpExtTCPAckCompressed" IpInReceives 123250 0.0 IpOutRequests 3684 0.0 TcpInSegs 123251 0.0 TcpOutSegs 3684 0.0 TcpExtTCPAckCompressed 119252 0.0 Signed-off-by:
-
Eric Dumazet authored
When TCP receives an out-of-order packet, it immediately sends a SACK packet, generating network load but also forcing the receiver to send 1-MSS pathological packets, increasing its RTX queue length/depth, and thus processing time. Wifi networks suffer from this aggressive behavior, but generally speaking, all these SACK packets add fuel to the fire when networks are under congestion. This patch adds a high resolution timer and tp->compressed_ack counter. Instead of sending a SACK, we program this timer with a small delay, based on RTT and capped to 1 ms : delay = min ( 5 % of RTT, 1 ms) If subsequent SACKs need to be sent while the timer has not yet expired, we simply increment tp->compressed_ack. When timer expires, a SACK is sent with the latest information. Whenever an ACK is sent (if data is sent, or if in-order data is received) timer is canceled. Note that tcp_sack_new_ofo_skb() is able to force a SACK to be sent if the sack blocks need to be shuffled, even if the timer has not expired. A new SNMP counter is added in the following patch. Two other patches add sysctls to allow changing the 1,000,000 and 44 values that this commit hard-coded. Signed-off-by:
Yuchung Cheng <ycheng@google.com> Acked-by:
Toke Høiland-Jørgensen <toke@toke.dk> Signed-off-by:
-
Eric Dumazet authored
As explained in commit 9f9843a7 ("tcp: properly handle stretch acks in slow start"), TCP stacks have to consider how many packets are acknowledged in one single ACK, because of GRO, but also because of ACK compression or losses. We plan to add SACK compression in the following patch, we must therefore not call tcp_enter_quickack_mode() Signed-off-by:
Soheil Hassas Yeganeh <soheil@google.com> Signed-off-by:
-
Eric Dumazet authored
Socket can not disappear under us. Signed-off-by:
-
Alexandre Belloni authored
ocelot_qsys.h is missing the SPDX identfier, fix that. Signed-off-by:
Alexandre Belloni <alexandre.belloni@bootlin.com> Reviewed-by:
Allan W. Nielsen <allan.nielsen@microsemi.com> Signed-off-by:
-
David S. Miller authored
Jose Abreu says: ==================== net: stmmac: Clean-up and tune-up This targets to uniformize the handling of the different GMAC versions in stmmac_main.c file and also tune-up the HW. Currently there are some if/else conditions in the main source file which calls different callbacks depending on the ID of GMAC. With the introducion of a generic HW interface handling which automatically selects the GMAC callbacks to be used, it is now unpleasant to see if conditions in the main code because this should be completely agnostic of the GMAC version. This series removes most of these conditions. There are some if conditions that remain untouched but the callbacks handling are now uniformized. Tested in GMAC5, hope I didn't break any previous versions. Please check [1] for performance analisys of patches 3-12. ==================== Signed-off-by: -
Jose Abreu authored
We can remove the if condition and check if return code is different than -EINVAL, meaning callback is present. Signed-off-by:
Jose Abreu <joabreu@synopsys.com> Cc: David S. Miller <davem@davemloft.net> Cc: Joao Pinto <jpinto@synopsys.com> Cc: Vitor Soares <soares@synopsys.com> Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com> Cc: Alexandre Torgue <alexandre.torgue@st.com> Signed-off-by:
-
Jose Abreu authored
Stop using if conditions depending on the GMAC version for getting the descriptor skbuff address and use instead a helper implemented in the descriptor files. Signed-off-by:
-
Jose Abreu authored
Currently an if condition is used to select the correct callback to set rx_onwer in descriptor. Lets keep this simple and always use the same callback. Signed-off-by:
-
Jose Abreu authored
We either have .enable_dma_transmission or .set_tx_tail_ptr in the HW table callbacks, we can never have both so there is no need to check for GMAC version. Signed-off-by:
-
Jose Abreu authored
Instead of relying on the GMAC version for choosing if we need to use dma_init or dma_init_{rx/tx}_chan callback, lets uniformize this and always use the dma_init_{rx/tx}_chan callbacks. While at it, fix the use of dma_init_chan callback, which shall be called for as many channels as the max of rx/tx channels. Signed-off-by: -
Jose Abreu authored
PTP and MMC modules base address can depend on the GMAC version. As this is HW specific lets move this base address calculation to hwif.c. Also, add an entry in the HW table so that we can specify the module offset. This can later be extended to more modules, if deemed necessary. Signed-off-by:
-
Jose Abreu authored
With the introducion of callbacks check in hwif.h we only call the callback if HW supports it so there is no longer need to check for GMAC version. Signed-off-by:
-
Jose Abreu authored
Instead of relying on the GMAC version for choosing if we need to use dma_{rx/tx}_mode or just dma_mode callback lets uniformize this and always use the dma_{rx/tx}_mode callbacks. Signed-off-by: -
Jose Abreu authored
Stop using if conditions depending on the GMAC version for clearing the descriptor and use instead a helper implemented in the descriptor files. Signed-off-by:
-
Jose Abreu authored
Stop using if conditions depending on the GMAC version for setting the the descriptor skbuff address and use instead a helper implemented in the descriptor files. Signed-off-by:
-
Jose Abreu authored
This is cutting down performance. Once the timer is armed it should run after the time expires for the first packet sent and not the last one. After this change, running iperf, the performance gain is +/- 24%. Signed-off-by:
-
Jose Abreu authored
This enables OSP (Operate on Second Packet) for GMAC4. The feature allows DMA to fetch second descriptor while its still processing the first one. Running iperf, the performance gain is +/- 38%. Signed-off-by:
-
- 17 May, 2018 18 commits
-
-
Shahar Klein authored
Set a specific source e-switch when setting a rule that matches on the ingress port. Signed-off-by:
Shahar Klein <shahark@mellanox.com> Reviewed-by:
Or Gerlitz <ogerlitz@mellanox.com> Reviewed-by:
Roi Dayan <roid@mellanox.com> Signed-off-by:
Saeed Mahameed <saeedm@mellanox.com>
-
Shahar Klein authored
The source e-switch owner allows a vport on one e-switch port be associated with a rule defined on the second port e-switch. The role of the source eswitch owner valid bit in the flow group is to allow the firmware fail driver attempts to wild card the source eswitch match field. If this bit is not set, the firmware ignores the source eswitch owner field totally. Signed-off-by:
-
Rabie Loulou authored
Set a specific destination e-switch when setting a destination vport. Signed-off-by:
Rabie Loulou <rabiel@mellanox.com> Reviewed-by:
-
Shahar Klein authored
The destination e-switch owner allows a rule in namespace of one e-switch owner to point to a vport that is natively associated with another e-switch owner. Signed-off-by:
-
Shahar Klein authored
When creating FTE, properly distinguish between destination being vport or tir. The previous code just worked accidentally b/c of both dest being in the same offset within a union. Signed-off-by:
-
Roi Dayan authored
When merged e-switch is supported, the per-port e-switch is logically merged into one e-switch that spans both physical ports and all the VFs. Under merged eswitch, both the matching on source vport and setting destination vport can have a 2nd attribute which is the vhca id of the eswitch owner. For example: esw0: {match: <src vport=1 owner=0> action: fwd to <dst vport=7, owner=1>} is a flow set on eswitch0 matching on source vport=1 from his eswitch and the action being fwd to dest vport=7 of eswitch1. Signed-off-by: -
David S. Miller authored
Florian Fainelli says: ==================== net: Allow more drivers with COMPILE_TEST This patch series includes more drivers to be build tested with COMPILE_TEST enabled. This helps cover some of the issues I just ran into with missing a driver *sigh*. Chanves in v3: - drop the TI Keystone NETCP driver from the COMPILE_TEST additions Changes in v2: - allow FEC to build outside of CONFIG_ARM/ARM64 by defining a layout of registers, this is not meant to run, so this is not a real issue if we are not matching the correct register layout ==================== Signed-off-by: -
Florian Fainelli authored
Those drivers build just fine with COMPILE_TEST, so make that possible. Signed-off-by:
Florian Fainelli <f.fainelli@gmail.com> Signed-off-by:
-
Florian Fainelli authored
The Freescale FEC driver builds fine with COMPILE_TEST, so make that possible. Signed-off-by:
-
Florian Fainelli authored
Most of the TI drivers build just fine with COMPILE_TEST, cpmac (AR7) is the exception because it uses a header file from arch/mips/include/asm/mach-ar7/ar7.h and keystone netcp which requires help from drivers/soc/ti/ for queue management helpers. Signed-off-by:
-
David Ahern authored
Add informative messages for error paths related to adding a VLAN to a device. Signed-off-by:
David Ahern <dsahern@gmail.com> Signed-off-by:
-
Manish Chopra authored
This patch makes use of build_skb() throughout in driver's receieve data path [HW gro flow and non HW gro flow]. With this, driver can build skb directly from the page segments which are already mapped to the hardware instead of allocating new SKB via netdev_alloc_skb() and memcpy the data which is quite costly. This really improves performance (keeping same or slight gain in rx throughput) in terms of CPU utilization which is significantly reduced [almost half] in non HW gro flow where for every incoming MTU sized packet driver had to allocate skb, memcpy headers etc. Additionally in that flow, it also gets rid of bunch of additional overheads [eth_get_headlen() etc.] to split headers and data in the skb. Tested with: system: 2 sockets, 4 cores per socket, hyperthreading, 2x4x2=16 cores iperf [server]: iperf -s iperf [client]: iperf -c <server_ip> -t 500 -i 10 -P 32 HW GRO off – w/o build_skb(), throughput: 36.8 Gbits/sec Average: CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle Average: all 0.59 0.00 32.93 0.00 0.00 43.07 0.00 0.00 23.42 HW GRO off - with build_skb(), throughput: 36.9 Gbits/sec Average: CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle Average: all 0.70 0.00 31.70 0.00 0.00 25.68 0.00 0.00 41.92 HW GRO on - w/o build_skb(), throughput: 36.9 Gbits/sec Average: CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle Average: all 0.86 0.00 24.14 0.00 0.00 6.59 0.00 0.00 68.41 HW GRO on - with build_skb(), throughput: 37.5 Gbits/sec Average: CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle Average: all 0.87 0.00 23.75 0.00 0.00 6.19 0.00 0.00 69.19 Signed-off-by:
Ariel Elior <ariel.elior@cavium.com> Signed-off-by:
Manish Chopra <manish.chopra@cavium.com> Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queueDavid S. Miller authored
Jeff Kirsher says: ==================== 10GbE Intel Wired LAN Driver Updates 2018-05-17 This series contains updates to ixgbe, ixgbevf and ice drivers. Cathy Zhou resolves sparse warnings by using the force attribute. Mauro S M Rodrigues fixes a bug where IRQs were not freed if a PCI error recovery system opts to remove the device which causes ixgbe_io_error_detected() to return PCI_ERS_RESULT_DISCONNECT before calling ixgbe_close_suspend() which results in IRQs not freed and crashing when the remove handler calls pci_disable_device(). Resolved this by calling ixgbe_close_suspend() before evaluating the PCI channel state. Pavel Tatashin releases the rtnl_lock during the call to ixgbe_close_suspend() to allow scaling if device_shutdown() is multi-threaded. Emil modifies ixgbe to not validate the MAC address during a reset, unless the MAC was set on the host so that the VF will get a new MAC address every time it reloads. Also updates ixgbevf to set hw->mac.perm_addr in order to retain the custom MAC on a reset. Anirudh updates the ice NVM read/erase/update AQ commands to align with the latest specification. ==================== Signed-off-by: -
Roman Mashak authored
Reported-by:
Vlad Buslov <vladbu@mellanox.com> Reported-by:
Davide Caratti <dcaratti@redhat.com> Signed-off-by:
Roman Mashak <mrv@mojatatu.com> Signed-off-by:
-
Dan Carpenter authored
We recently refactored this code and introduced a static checker warning. Smatch complains that if cmd->index is zero then we would underflow the arrays. That's obviously true. The question is whether we prevent cmd->index from being zero at a different level. I've looked at the code and I don't immediately see a check for that. Fixes: 062b3e1b ("net/ncsi: Refactor MAC, VLAN filters") Signed-off-by:
Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by:
-
Eric Dumazet authored
syzkaller found that following program crashes the host : { int fd = socket(AF_SMC, SOCK_STREAM, 0); int val = 1; listen(fd, 0); shutdown(fd, SHUT_RDWR); setsockopt(fd, 6, TCP_NODELAY, &val, 4); } Simply initialize conn.tx_work & conn.send_lock at socket creation, rather than deeper in the stack. ODEBUG: assert_init not available (active state 0) object type: timer_list hint: (null) WARNING: CPU: 1 PID: 13988 at lib/debugobjects.c:329 debug_print_object+0x16a/0x210 lib/debugobjects.c:326 Kernel panic - not syncing: panic_on_warn set ... CPU: 1 PID: 13988 Comm: syz-executor0 Not tainted 4.17.0-rc4+ #46 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 Call Trace: __dump_stack lib/dump_stack.c:77 [inline] dump_stack+0x1b9/0x294 lib/dump_stack.c:113 panic+0x22f/0x4de kernel/panic.c:184 __warn.cold.8+0x163/0x1b3 kernel/panic.c:536 report_bug+0x252/0x2d0 lib/bug.c:186 fixup_bug arch/x86/kernel/traps.c:178 [inline] do_error_trap+0x1de/0x490 arch/x86/kernel/traps.c:296 do_invalid_op+0x1b/0x20 arch/x86/kernel/traps.c:315 invalid_op+0x14/0x20 arch/x86/entry/entry_64.S:992 RIP: 0010:debug_print_object+0x16a/0x210 lib/debugobjects.c:326 RSP: 0018:ffff880197a37880 EFLAGS: 00010086 RAX: 0000000000000061 RBX: 0000000000000005 RCX: ffffc90001ed0000 RDX: 0000000000004aaf RSI: ffffffff8160f6f1 RDI: 0000000000000001 RBP: ffff880197a378c0 R08: ffff8801aa7a0080 R09: ffffed003b5e3eb2 R10: ffffed003b5e3eb2 R11: ffff8801daf1f597 R12: 0000000000000001 R13: ffffffff88d96980 R14: ffffffff87fa19a0 R15: ffffffff81666ec0 debug_object_assert_init+0x309/0x500 lib/debugobjects.c:692 debug_timer_assert_init kernel/time/timer.c:724 [inline] debug_assert_init kernel/time/timer.c:776 [inline] del_timer+0x74/0x140 kernel/time/timer.c:1198 try_to_grab_pending+0x439/0x9a0 kernel/workqueue.c:1223 mod_delayed_work_on+0x91/0x250 kernel/workqueue.c:1592 mod_delayed_work include/linux/workqueue.h:541 [inline] smc_setsockopt+0x387/0x6d0 net/smc/af_smc.c:1367 __sys_setsockopt+0x1bd/0x390 net/socket.c:1903 __do_sys_setsockopt net/socket.c:1914 [inline] __se_sys_setsockopt net/socket.c:1911 [inline] __x64_sys_setsockopt+0xbe/0x150 net/socket.c:1911 do_syscall_64+0x1b1/0x800 arch/x86/entry/common.c:287 entry_SYSCALL_64_after_hwframe+0x49/0xbe Fixes: 01d2f7e2 ("net/smc: sockopts TCP_NODELAY and TCP_CORK") Signed-off-by:syzbot <syzkaller@googlegroups.com> Signed-off-by:
-
Jiri Pirko authored
Don't store repr pointer to reprs array until the representor is successfully created. This avoids message about "representor destruction" even when it was never created. Also it cleans-up the flow. Also, check return value after port alloc. Signed-off-by:
Jiri Pirko <jiri@mellanox.com> Reviewed-by:
Jakub Kicinski <jakub.kicinski@netronome.com> Signed-off-by:
-
David S. Miller authored
Antoine Tenart says: ==================== net: mvpp2: small improvements Those 3 patches are small improvements to the Marvell PPv2 driver. The series does not conflict with the one sent about phylink and 1000/2500baseX support, so the two series can live in parallel. ==================== Signed-off-by:
-
